基于LSTM-Transformer-ECA的海上风电施工船舶轨迹预测方法-《集美大学学报（自然科学版）》

文章信息/Info

Title:: LSTM-Transformer-ECA Method for Offshore Wind Construction Vessel Trajectory Prediction

Author(s):: HU Penghui; CHEN Jinhai; (Navigation College of Jimei University,Xiamen 361021,China)

Keywords:: offshore wind power; construction vessels; trajectory prediction; LSTM; Transformer

摘要:: 提出了一种融合长短期记忆（LSTM）网络和Transformer编码器的预测方法，进一步集成了增强型通道注意力机制，以深入挖掘船舶航行的内在特征。首先对船舶自动识别系统的数据进行清洗、插值处理、重采样以及标准化，以获取精确的船舶航行路径。接着，建立了一个结合LSTM-Transformer-ECA的复合神经网络模型，并完成了模型参数的设定。最后，用施工船舶的AIS数据进行实验测试，并与LSTM、Bi-LSTM、Transformer进行了比较分析。实验结果表明，用本文提出的方法预测施工船舶航迹的经度和纬度，平均绝对误差和均方根误差两个指标改进显著。

Abstract:: This paper introduces a forecasting method that integrates long short-term memory (LSTM) networks with Transformer encoders,further enhanced by the incorporation of an enhanced channel attention (ECA) mechanism to deeply explore the intrinsic characteristics of vessel navigation.Initially,the paper undertakes the cleaning,interpolation,resampling,and normalization of automatic identification system (AIS) data to accurately capture the navigation paths of vessels.Subsequently,a composite neural network model that combines LSTM,Transformer,and ECA is established,with the model parameters being appropriately initialized.Finally,experimental validation is conducted using AIS data from construction vessels,and a comparative analysis is performed with models that utilize LSTM,Transformer,and Bi-LSTM.The experimental results demonstrate that the method proposed in this paper significantly improves the precision in the prediction of longitude and latitude,as evidenced by a notable reduction in both the mean absolute error (MAE) and the root mean square error (RMSE) metrics.